* [PATCH V2 2/3] nvmet-rdma: Implement get_mdts controller op
2020-03-05 9:55 [PATCH V2 1/3] nvmet: Add get_mdts op for controllers Max Gurtovoy
@ 2020-03-05 9:55 ` Max Gurtovoy
2020-03-05 20:44 ` Sagi Grimberg
2020-03-05 9:55 ` [PATCH V2 3/3] nvmet-rdma: allocate RW ctxs according to mdts Max Gurtovoy
2020-03-05 20:40 ` [PATCH V2 1/3] nvmet: Add get_mdts op for controllers Sagi Grimberg
2 siblings, 1 reply; 9+ messages in thread
From: Max Gurtovoy @ 2020-03-05 9:55 UTC (permalink / raw)
To: jgg, linux-nvme, sagi, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, Max Gurtovoy, bharat, nirranjan, bvanassche
Set the maximal data transfer size to be 1MB (currently mdts is
unlimited). This will allow calculating the amount of MR's that
one ctrl should allocate to fulfill it's capabilities.
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
---
changes from V1:
- renamed nvmet_rdma_set_mdts to nvmet_rdma_get_mdts
- align to get_mdts callback changes
---
drivers/nvme/target/rdma.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index 37d262a..d12ef0d 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -30,6 +30,7 @@
#define NVMET_RDMA_DEFAULT_INLINE_DATA_SIZE PAGE_SIZE
#define NVMET_RDMA_MAX_INLINE_SGE 4
#define NVMET_RDMA_MAX_INLINE_DATA_SIZE max_t(int, SZ_16K, PAGE_SIZE)
+#define NVMET_RDMA_MAX_MDTS 8
struct nvmet_rdma_cmd {
struct ib_sge sge[NVMET_RDMA_MAX_INLINE_SGE + 1];
@@ -1602,6 +1603,12 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req,
}
}
+static u8 nvmet_rdma_get_mdts(const struct nvmet_ctrl *ctrl)
+{
+ /* Assume mpsmin == device_page_size == 4KB */
+ return NVMET_RDMA_MAX_MDTS;
+}
+
static const struct nvmet_fabrics_ops nvmet_rdma_ops = {
.owner = THIS_MODULE,
.type = NVMF_TRTYPE_RDMA,
@@ -1612,6 +1619,7 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req,
.queue_response = nvmet_rdma_queue_response,
.delete_ctrl = nvmet_rdma_delete_ctrl,
.disc_traddr = nvmet_rdma_disc_port_addr,
+ .get_mdts = nvmet_rdma_get_mdts,
};
static void nvmet_rdma_remove_one(struct ib_device *ib_device, void *client_data)
--
1.8.3.1
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply related [flat|nested] 9+ messages in thread* Re: [PATCH V2 2/3] nvmet-rdma: Implement get_mdts controller op
2020-03-05 9:55 ` [PATCH V2 2/3] nvmet-rdma: Implement get_mdts controller op Max Gurtovoy
@ 2020-03-05 20:44 ` Sagi Grimberg
2020-03-05 23:01 ` Max Gurtovoy
0 siblings, 1 reply; 9+ messages in thread
From: Sagi Grimberg @ 2020-03-05 20:44 UTC (permalink / raw)
To: Max Gurtovoy, jgg, linux-nvme, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, bharat, nirranjan, bvanassche
On 3/5/20 1:55 AM, Max Gurtovoy wrote:
> Set the maximal data transfer size to be 1MB (currently mdts is
> unlimited). This will allow calculating the amount of MR's that
> one ctrl should allocate to fulfill it's capabilities.
>
> Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
> ---
>
> changes from V1:
> - renamed nvmet_rdma_set_mdts to nvmet_rdma_get_mdts
> - align to get_mdts callback changes
>
> ---
> drivers/nvme/target/rdma.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
> index 37d262a..d12ef0d 100644
> --- a/drivers/nvme/target/rdma.c
> +++ b/drivers/nvme/target/rdma.c
> @@ -30,6 +30,7 @@
> #define NVMET_RDMA_DEFAULT_INLINE_DATA_SIZE PAGE_SIZE
> #define NVMET_RDMA_MAX_INLINE_SGE 4
> #define NVMET_RDMA_MAX_INLINE_DATA_SIZE max_t(int, SZ_16K, PAGE_SIZE)
> +#define NVMET_RDMA_MAX_MDTS 8
>
> struct nvmet_rdma_cmd {
> struct ib_sge sge[NVMET_RDMA_MAX_INLINE_SGE + 1];
> @@ -1602,6 +1603,12 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req,
> }
> }
>
> +static u8 nvmet_rdma_get_mdts(const struct nvmet_ctrl *ctrl)
> +{
> + /* Assume mpsmin == device_page_size == 4KB */
This comment should come at the define entry.
> + return NVMET_RDMA_MAX_MDTS;
> +}
> +
> static const struct nvmet_fabrics_ops nvmet_rdma_ops = {
> .owner = THIS_MODULE,
> .type = NVMF_TRTYPE_RDMA,
> @@ -1612,6 +1619,7 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req,
> .queue_response = nvmet_rdma_queue_response,
> .delete_ctrl = nvmet_rdma_delete_ctrl,
> .disc_traddr = nvmet_rdma_disc_port_addr,
> + .get_mdts = nvmet_rdma_get_mdts,
> };
>
> static void nvmet_rdma_remove_one(struct ib_device *ib_device, void *client_data)
>
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH V2 2/3] nvmet-rdma: Implement get_mdts controller op
2020-03-05 20:44 ` Sagi Grimberg
@ 2020-03-05 23:01 ` Max Gurtovoy
0 siblings, 0 replies; 9+ messages in thread
From: Max Gurtovoy @ 2020-03-05 23:01 UTC (permalink / raw)
To: Sagi Grimberg, jgg, linux-nvme, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, bharat, nirranjan, bvanassche
On 3/5/2020 10:44 PM, Sagi Grimberg wrote:
>
>
> On 3/5/20 1:55 AM, Max Gurtovoy wrote:
>> Set the maximal data transfer size to be 1MB (currently mdts is
>> unlimited). This will allow calculating the amount of MR's that
>> one ctrl should allocate to fulfill it's capabilities.
>>
>> Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
>> ---
>>
>> changes from V1:
>> - renamed nvmet_rdma_set_mdts to nvmet_rdma_get_mdts
>> - align to get_mdts callback changes
>>
>> ---
>> drivers/nvme/target/rdma.c | 8 ++++++++
>> 1 file changed, 8 insertions(+)
>>
>> diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
>> index 37d262a..d12ef0d 100644
>> --- a/drivers/nvme/target/rdma.c
>> +++ b/drivers/nvme/target/rdma.c
>> @@ -30,6 +30,7 @@
>> #define NVMET_RDMA_DEFAULT_INLINE_DATA_SIZE PAGE_SIZE
>> #define NVMET_RDMA_MAX_INLINE_SGE 4
>> #define NVMET_RDMA_MAX_INLINE_DATA_SIZE max_t(int, SZ_16K,
>> PAGE_SIZE)
>> +#define NVMET_RDMA_MAX_MDTS 8
>> struct nvmet_rdma_cmd {
>> struct ib_sge sge[NVMET_RDMA_MAX_INLINE_SGE + 1];
>> @@ -1602,6 +1603,12 @@ static void nvmet_rdma_disc_port_addr(struct
>> nvmet_req *req,
>> }
>> }
>> +static u8 nvmet_rdma_get_mdts(const struct nvmet_ctrl *ctrl)
>> +{
>> + /* Assume mpsmin == device_page_size == 4KB */
>
> This comment should come at the define entry.
Sure, no problem.
>
>> + return NVMET_RDMA_MAX_MDTS;
>> +}
>> +
>> static const struct nvmet_fabrics_ops nvmet_rdma_ops = {
>> .owner = THIS_MODULE,
>> .type = NVMF_TRTYPE_RDMA,
>> @@ -1612,6 +1619,7 @@ static void nvmet_rdma_disc_port_addr(struct
>> nvmet_req *req,
>> .queue_response = nvmet_rdma_queue_response,
>> .delete_ctrl = nvmet_rdma_delete_ctrl,
>> .disc_traddr = nvmet_rdma_disc_port_addr,
>> + .get_mdts = nvmet_rdma_get_mdts,
>> };
>> static void nvmet_rdma_remove_one(struct ib_device *ib_device,
>> void *client_data)
>>
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH V2 3/3] nvmet-rdma: allocate RW ctxs according to mdts
2020-03-05 9:55 [PATCH V2 1/3] nvmet: Add get_mdts op for controllers Max Gurtovoy
2020-03-05 9:55 ` [PATCH V2 2/3] nvmet-rdma: Implement get_mdts controller op Max Gurtovoy
@ 2020-03-05 9:55 ` Max Gurtovoy
2020-03-05 20:49 ` Sagi Grimberg
2020-03-05 20:40 ` [PATCH V2 1/3] nvmet: Add get_mdts op for controllers Sagi Grimberg
2 siblings, 1 reply; 9+ messages in thread
From: Max Gurtovoy @ 2020-03-05 9:55 UTC (permalink / raw)
To: jgg, linux-nvme, sagi, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, Max Gurtovoy, bharat, nirranjan, bvanassche
Current nvmet-rdma code allocates MR pool budget based on queue size,
assuming both host and target use the same "max_pages_per_mr" count.
After limiting the mdts value for RDMA controllers, we know the factor
of maximum MR's per IO operation. Thus, make sure MR pool will be
sufficient for the required IO depth and IO size.
That is, say host's SQ size is 100, then the MR pool budget allocated
currently at target will also be 100 MRs. But 100 IO WRITE Requests
with 256 sg_count(IO size above 1MB) require 200 MRs when target's
"max_pages_per_mr" is 128.
Reported-by: Krishnamraju Eraparaju <krishna2@chelsio.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
---
changes from V1:
- added "Reviewed-by" signature (Christoph)
---
drivers/nvme/target/rdma.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index d12ef0d..daab656 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -976,7 +976,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
{
struct ib_qp_init_attr qp_attr;
struct nvmet_rdma_device *ndev = queue->dev;
- int comp_vector, nr_cqe, ret, i;
+ int comp_vector, nr_cqe, ret, i, factor;
/*
* Spread the io queues across completion vectors,
@@ -1009,7 +1009,9 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
qp_attr.qp_type = IB_QPT_RC;
/* +1 for drain */
qp_attr.cap.max_send_wr = queue->send_queue_size + 1;
- qp_attr.cap.max_rdma_ctxs = queue->send_queue_size;
+ factor = rdma_rw_mr_factor(ndev->device, queue->cm_id->port_num,
+ 1 << NVMET_RDMA_MAX_MDTS);
+ qp_attr.cap.max_rdma_ctxs = queue->send_queue_size * factor;
qp_attr.cap.max_send_sge = max(ndev->device->attrs.max_sge_rd,
ndev->device->attrs.max_send_sge);
--
1.8.3.1
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply related [flat|nested] 9+ messages in thread* Re: [PATCH V2 3/3] nvmet-rdma: allocate RW ctxs according to mdts
2020-03-05 9:55 ` [PATCH V2 3/3] nvmet-rdma: allocate RW ctxs according to mdts Max Gurtovoy
@ 2020-03-05 20:49 ` Sagi Grimberg
2020-03-05 22:59 ` Max Gurtovoy
0 siblings, 1 reply; 9+ messages in thread
From: Sagi Grimberg @ 2020-03-05 20:49 UTC (permalink / raw)
To: Max Gurtovoy, jgg, linux-nvme, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, bharat, nirranjan, bvanassche
> diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
> index d12ef0d..daab656 100644
> --- a/drivers/nvme/target/rdma.c
> +++ b/drivers/nvme/target/rdma.c
> @@ -976,7 +976,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
> {
> struct ib_qp_init_attr qp_attr;
> struct nvmet_rdma_device *ndev = queue->dev;
> - int comp_vector, nr_cqe, ret, i;
> + int comp_vector, nr_cqe, ret, i, factor;
>
> /*
> * Spread the io queues across completion vectors,
> @@ -1009,7 +1009,9 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
> qp_attr.qp_type = IB_QPT_RC;
> /* +1 for drain */
> qp_attr.cap.max_send_wr = queue->send_queue_size + 1;
> - qp_attr.cap.max_rdma_ctxs = queue->send_queue_size;
> + factor = rdma_rw_mr_factor(ndev->device, queue->cm_id->port_num,
> + 1 << NVMET_RDMA_MAX_MDTS);
Maybe I'm missing something, but aren't you missing the mpsmin
multiplier? your maxpages is not (1 << 8) but rather (1 << 20) isn't it?
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH V2 3/3] nvmet-rdma: allocate RW ctxs according to mdts
2020-03-05 20:49 ` Sagi Grimberg
@ 2020-03-05 22:59 ` Max Gurtovoy
2020-03-06 1:19 ` Sagi Grimberg
0 siblings, 1 reply; 9+ messages in thread
From: Max Gurtovoy @ 2020-03-05 22:59 UTC (permalink / raw)
To: Sagi Grimberg, jgg, linux-nvme, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, bharat, nirranjan, bvanassche
On 3/5/2020 10:49 PM, Sagi Grimberg wrote:
>
>> diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
>> index d12ef0d..daab656 100644
>> --- a/drivers/nvme/target/rdma.c
>> +++ b/drivers/nvme/target/rdma.c
>> @@ -976,7 +976,7 @@ static int nvmet_rdma_create_queue_ib(struct
>> nvmet_rdma_queue *queue)
>> {
>> struct ib_qp_init_attr qp_attr;
>> struct nvmet_rdma_device *ndev = queue->dev;
>> - int comp_vector, nr_cqe, ret, i;
>> + int comp_vector, nr_cqe, ret, i, factor;
>> /*
>> * Spread the io queues across completion vectors,
>> @@ -1009,7 +1009,9 @@ static int nvmet_rdma_create_queue_ib(struct
>> nvmet_rdma_queue *queue)
>> qp_attr.qp_type = IB_QPT_RC;
>> /* +1 for drain */
>> qp_attr.cap.max_send_wr = queue->send_queue_size + 1;
>> - qp_attr.cap.max_rdma_ctxs = queue->send_queue_size;
>> + factor = rdma_rw_mr_factor(ndev->device, queue->cm_id->port_num,
>> + 1 << NVMET_RDMA_MAX_MDTS);
>
> Maybe I'm missing something, but aren't you missing the mpsmin
> multiplier? your maxpages is not (1 << 8) but rather (1 << 20) isn't it?
why ?
we support 256 pages with 4KB size each to get 1MB "mdts".
The factor is in units of RW api context (that might limit the size of
the MRs according to some logic).
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH V2 3/3] nvmet-rdma: allocate RW ctxs according to mdts
2020-03-05 22:59 ` Max Gurtovoy
@ 2020-03-06 1:19 ` Sagi Grimberg
0 siblings, 0 replies; 9+ messages in thread
From: Sagi Grimberg @ 2020-03-06 1:19 UTC (permalink / raw)
To: Max Gurtovoy, jgg, linux-nvme, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, bharat, nirranjan, bvanassche
>>> diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
>>> index d12ef0d..daab656 100644
>>> --- a/drivers/nvme/target/rdma.c
>>> +++ b/drivers/nvme/target/rdma.c
>>> @@ -976,7 +976,7 @@ static int nvmet_rdma_create_queue_ib(struct
>>> nvmet_rdma_queue *queue)
>>> {
>>> struct ib_qp_init_attr qp_attr;
>>> struct nvmet_rdma_device *ndev = queue->dev;
>>> - int comp_vector, nr_cqe, ret, i;
>>> + int comp_vector, nr_cqe, ret, i, factor;
>>> /*
>>> * Spread the io queues across completion vectors,
>>> @@ -1009,7 +1009,9 @@ static int nvmet_rdma_create_queue_ib(struct
>>> nvmet_rdma_queue *queue)
>>> qp_attr.qp_type = IB_QPT_RC;
>>> /* +1 for drain */
>>> qp_attr.cap.max_send_wr = queue->send_queue_size + 1;
>>> - qp_attr.cap.max_rdma_ctxs = queue->send_queue_size;
>>> + factor = rdma_rw_mr_factor(ndev->device, queue->cm_id->port_num,
>>> + 1 << NVMET_RDMA_MAX_MDTS);
>>
>> Maybe I'm missing something, but aren't you missing the mpsmin
>> multiplier? your maxpages is not (1 << 8) but rather (1 << 20) isn't it?
>
> why ?
>
> we support 256 pages with 4KB size each to get 1MB "mdts".
>
> The factor is in units of RW api context (that might limit the size of
> the MRs according to some logic).
You're right, never mind...
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH V2 1/3] nvmet: Add get_mdts op for controllers
2020-03-05 9:55 [PATCH V2 1/3] nvmet: Add get_mdts op for controllers Max Gurtovoy
2020-03-05 9:55 ` [PATCH V2 2/3] nvmet-rdma: Implement get_mdts controller op Max Gurtovoy
2020-03-05 9:55 ` [PATCH V2 3/3] nvmet-rdma: allocate RW ctxs according to mdts Max Gurtovoy
@ 2020-03-05 20:40 ` Sagi Grimberg
2 siblings, 0 replies; 9+ messages in thread
From: Sagi Grimberg @ 2020-03-05 20:40 UTC (permalink / raw)
To: Max Gurtovoy, jgg, linux-nvme, hch, kbusch, Chaitanya.Kulkarni
Cc: krishna2, bharat, nirranjan, bvanassche
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
_______________________________________________
linux-nvme mailing list
linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 9+ messages in thread